Analog VLSI hardware for fuzzy systems

Our world has an analog nature and it is natural to process signals in an analog way. Analog signal processing can be much faster than digital signal processing and AD or DA conversion is not required. The main obstacle is to develop adequate circuits for nonlinear signal processing. In the paper several new circuits are proposed. These circuits use nonlinear characteristics of MOS transistors for nonlinear signal processing. The fuzzy signal processing is used as an example. The proposed fuzzifier circuits are relatively simple while almost arbitrary shapes of membership functions can be obtained. The proposed current mode MAX and MIN operators exhibit accuracy superior to other circuits. The defuzzifier circuit uses the concept of signal normalization and weighted sum. A new normalization circuit, operating in the subthreshold conduction mode, exhibits almost ideal characteristics. The described new building blocks were used to design the entire analog fuzzy VLSI chip